Application of QSP modeling to possible pharmacological targets in - - PowerPoint PPT Presentation
Application of QSP modeling to possible pharmacological targets in Alzheimers disease Oleg Demin, Tatiana Karelina, Sergey Belykh, Oleg Demin Jr Institute for Systems Biology SPb, Moscow, Russia Timothy Nicholas, Hugh A. Barton, Yasong Lu,
Oleg Demin, Tatiana Karelina, Sergey Belykh, Oleg Demin Jr
Institute for Systems Biology SPb, Moscow, Russia
Loughborough, September 7, 2012
Timothy Nicholas, Hugh A. Barton, Yasong Lu, Sridhar Duvvuri
Pfizer Global Research and Development, Groton, CT USA
Systems biology Systems pharmacology PK/PD modelling
Typically intracellular or molecular level Conceptually developed at the scale of interest, but is usually at a tissue/organ scale to provide compatibility with PKPD models Typically developed at the organism scale reflecting typical data collected Highly granular typically all network intermediates and their mechanistic linkage are captured. (network-level) Intermediate in granularity; typically considers key network intermediates; may contain mechanistic and empiric relationships between intermediates (pathway level) Low granularity; typically no further layer of complexity than observation (input/output-type model; target level) Assumption rich, typically need experimental data to calibrate Assumption-rich; may need experimental in vitro data to calibrate Low in assumptions; no further in vitro data required to calibrate
Systems Pharmacology Modeling: definition and comparison with other modeling techniques
Agoram B, Demin O; Drug Discov Today, 2011 Dec;16(23-24):1031-6.
regulatory mechanisms of disease development/progression and mechanisms of drug action at intracellular/tissue/organism levels.
drug pharmacokinetics, tissue cross-talks and allows to express intracellular effects of the drug in terms of clinically measured biomarkers and end-points.
Agoram BM, Demin O; Drug Discov Today, 2011 Dec;16(23-24):1031-6.
literature and input from disease biology experts an initial mathematical model can be constructed (Iteration 1).
subject to feedback from a wider expert panel and subsequently this iteration (Iteration 2) is used to identify critical assumptions and testable hypotheses. At this stage the determination of key system parameters such as target molecule concentrations (biomeasures) are likely to be critical.
produced that is consistent with these data (Iteration 3). This model can be used to predict clinical outcome and contribute to trial design eg initial dose
result can be compared to predictions and the conclusions incorporated in subsequent rounds
Benson N, Cucurull-Sanchez L, Demin O, Smirnov S, van der Graaf P.; Adv Exp Med Biol., 2012;736:607-15
project risks
project risks
cure for the disease.
that the disease is associated with senile plaques (extracellular deposits of amyloid) and neurofibrillary tangles (intracellular aggregates of hyperphosphorylated tau protein) in the brain.
Abeta leads to neurodegeneration (Hardy 1992, 2002, 2009). This toxic species are related to either soluble monomeric or soluble oligomeric Abeta. Abeta plaques are in equilibrium with the soluble ‘toxic’ species of Abeta.
with no observed cognitive benefit.
clinical trial for the duration necessary. QUESTION: How to estimate optimal potency/exposure of the modulators for Abeta production/clearance? ANSWER: To develop QSP model of Abeta aggregation in human
(1) in vivo data obtained by post-mortem autopsy for healthy subjects and for AD patients (Ab40, Ab42: soluble and unsoluble measured in nM) (2) Abeta production/clearance ration measured for AD patients in vivo (3) Positron Emission Tomography (PET) data measured in vivo for for healthy subjects and for AD patients
Positron Emission Tomography Positron Emission Tomography (PET) provides in vivo evidence of the local accumulation of a specific radioligand and allows to estimate quantitative features of the binding sites (their number and affinity) for this radioligand. In the case of Alzheimer Disease there is radioligand PIB that could bind with Amyloid Fibrils (with some motifs on fibril). Thus, this method allows to estimate quantity of monomers in all fibrils in vivo. Standardized Uptake Value [SUV] and Distribution Volume [DV] of radioligand for particular region of the brain or Ratio of Standardized Uptake Values [SUVR] and Ratio of Distribution Volumes [DVR] of 2 regions in the brain are the main parameters that represented the results of PET.
How to understand/express PET data in terms of Abeta aggregation mechanisms?
verify it against postmortem data
validate the model against available PET data measured for both healthy subjects and AD patients
pharmacological interventions directed toward
initiated at 40, 50 or 65 years of age, were performed.
Assumptions that we used to create the model:
different.
and fibrils breakage have equal rate constants.
law.
Main processes of Ab aggregation and oligomerization
Pi – Ab forms: P1 – monomer P2…Pn-1 – different oligomers Pn – nucleus (= fibril with i=n) Pj, where j = from (n+1) to infinity – fibrils P1 P1 P2 P3 P1 P1 ……… Pn-1 Oligomerization: n*P1 Pn Nucleation: Pn P1 Pn+1 Pn+2 P1 ……… infinity Fibrillation: Fibrils breakage: Pj
Fibril + Fibril Fibril + Oligomer Oligomer + Oligomer
P1 Pn Pn-1 Our model is based on 3 different models from [Science. 2009 Dec 11;326(5959):1533-7](this model includes only fibrils), [Nucl Med Biol. 2005 May;32(4):337-51](this model includes monomers and fibrils, also in this model PET is described) and [Biophys J. 2007 May 15;92(10):3448-58](this model includes oligomerization and fibrillation).
Ab40 monomer Ab40 Nucleus Ab40 monomer Ab40 Fibril Ab40 dimer Ab40 trimer Ab40 tetramer Ab42 monomer Ab42 Nucleus Ab42 monomer Ab42 Fibril Ab42 dimer Ab42 trimer Ab42 tetramer HeteroDimer HeteroTetramer
Scheme of aggregation model
Synthesis Degradation, Elimination Oligomerization and Aggregation processes
Ab40 Fibril Ab40 Fibril Ab42 monomer
Ab42 Fibril Ab42 Fibril Ab42 Fibril Ab40 Fibril Ab40 monomer
40 break
k
40
K
40 deg
k
40 syn
k
42 pol
k
40 pol
k
42 syn
k
42 deg
k
40
K
40
K
42
K
42
K
42
K
42 break
k
42 break
k
40 break
k
el
k
40 _ f nuc
k
40 _ r nuc
k
42 _ r nuc
k
42 _ f nuc
k
dim
K
tet
K
el
k
el
k
el
k
el
k
el
k
el
k
el
k
40 des
k
42 des
k
40 des
k
42 des
k
1 1 1 1
2 ) 1 ( 2 2
i j j break i break i pol i pol i des i
P k P i k P P k P P k P k dt dP
Aggregation model for healthy subjects Because we don’t know the maximal number of monomers in fibril, we assume it equal to
with fibril of i-length: destruction (degradation), forming of fibril of i-length by binding of monomer to fibril of (i-1)-length, forming of (i+1)-length by binding of monomer to fibril of i- length, breakage of fibril of i-length, formation of fibril of i-length by breakage of longer fibrils: We sum up this infinite system of differential equations using the method described in [Science. 2009 Dec 11;326(5959):1533-7]:
40 40 40 40 40 40 1 40 40 40 40
3 2 F k Fm k N P k F k dt dF
break break pol des
40 40 40 40 1 40 40 40 1 40 40 40 40
2 2 4 F k F P k N P k Fm k dt dFm
break pol pol des
40 40 1 40 40 40 _ 40 1 40 _ 40 40 40
2 N P k N k P k N k dt dN
pol r nuc f nuc des
) 40 ( ) ( 2 2 40
40 1 40 40 40 1 40 40 40 40 40 _ 40 1 40 _ 40 1 40 deg 40
P S k F N P k F k N k P k P k k dt dS
el pol break r nuc f nuc syn
1 n i i
P F
1 n i i
P i Fm
Soluble Ab40: S40 = P1 + 2*P2 + 3*P3 + 4*P4 + D + 2*T; N40 - Ab40 nucleus Assumption: nucleation is reversible and nucleus contains one monomer ([Biophys J. 2007 May 15;92(10):3448-58], [Nucl Med Biol. 2005 May;32(4):337-51]).
Model consists of 8 ODEs (Ab40 and Ab42) and includes 19 parameters. 5 parameters have been estimated on the basis of allometric scaling of “rat”
been fitted to human in vivo data
Human model verification [HEALTHY SUBJECTS] Verification of model parameters for human against data obtained by post-mortem autopsy for healthy subjects and for AD patients. Because deposition of Ab appears in different part of human brain in different time, for fitting we use only data on Ab concentration in different regions of cortex, because in almost all regions of cortex Ab begin to deposit at the first or second stages of AD [J Cell Mol Med. 2008 Oct;12(5B):1848-62]. First of all we verify our model against data for healthy subjects:
References: Exp Neurol. 1999 Aug;158(2):328-37 [10] Arch Neurol. 2009 Feb;66(2):190-9 [7] Arch Neurol. 2008 Jul;65(7):906-12 [13] References: Black point: Exp Neurol. 1999 Aug;158(2):328-37 [10] Arch Neurol. 2009 Feb;66(2):190-9 [7] Arch Neurol. 2008 Jul;65(7):906-12 [13] Blue point: Brain Pathol. 2010 Jul;20(4):787-93 [individual]
Soluble Ab40 and Ab42
Human model verification [HEALTHY SUBJECTS] Verification of model parameters for human against data obtained by post-mortem autopsy for healthy subjects and for AD patients. Because deposition of Ab appears in different part of human brain in different time, for fitting we use only data on Ab concentration in different regions of cortex, because in almost all regions of cortex Ab begin to deposit at the first or second stages of AD [J Cell Mol Med. 2008 Oct;12(5B):1848-62]. First of all we verify our model against data for healthy subjects:
References: Black point: Am J Pathol. 1998 Jun;152(6):1633-40 [individual] Blue point: Arch Neurol. 2009 Feb;66(2):190-9 [10] + Arch Neurol. 2008 Jul;65(7):906-12 [7] + Exp Neurol. 1999 Aug;158(2):328-37 [13] Red point: Neurobiol Aging. 2008 Feb;29(2):210-21 [individual] Green point: Am J Pathol. 1998 Jun;152(6):1633-40 [individual] Pink point: Am J Pathol. 2000 Dec;157(6):2093-9 [individual]
Insoluble Ab40 and Ab42 Logarithmic scale Logarithmic scale
From Healthy subject to AD patient In [Science. 2010 Dec 24;330(6012):1774. Epub 2010 Dec 9] it was shown that synthesis to clearance ratio is increased by 30% at about 75 years in AD patient in comparison with healthy subjects. This means that if synthesis rate is constant, clearance should be reduced by 23%. So we add in the model function that will be decreased in such manner starting from 40 years: AD1=(1+k_chg*Time/Km)/(1+Time/Km); Where k_chg = 0.73, Km = 5;
Human model validation for AD patients Using this function we simulate Soluble and Insoluble Ab and compare it with experimental data for AD patients. Soluble Ab40 and Ab42 Insoluble Ab40 and Ab42
Black point: Exp Neurol. 1999 Aug;158(2):328-37 [23] + Arch Neurol. 2007 Mar;64(3):431-4 [1] Blue point: Arch Neurol. 2009 Feb;66(2):190-9 [9] Red and Green point: Arch Neurol. 2008 Jul;65(7):906-12 [27]
Black point: Exp Neurol. 1999 Aug;158(2):328-37 [23] + Arch Neurol. 2007 Mar;64(3):431-4 [1] Blue point: Arch Neurol. 2009 Feb;66(2):190-9 [9] Red point: Am J Pathol. 1998 Jun;152(6):1633-40 [individual] Green point: Arch Neurol. 2008 Jul;65(7):906-12 [27] Pink point: Arch Neurol. 2008 Jul;65(7):906-12 [27]
It is not fitting It is not fitting It is not fitting It is not fitting zoom
Black point: Exp Neurol. 1999 Aug;158(2):328-37 [23] + Arch Neurol. 2007 Mar;64(3):431-4 [1] Blue point: Arch Neurol. 2009 Feb;66(2):190-9 [9] Red and Green point: Arch Neurol. 2008 Jul;65(7):906-12 [27]
How to describe PET data in terms of variables of model developed?
bind with Amyloid Fibrils (with some motifs on fibril). PET allows to estimate quantity of monomers in all fibrils in vivo.
regions in the brain are the main parameters that represented the results of PET.
amyloid in different regions of the brain could be measured (Cpet).
amyloid DVR and SUVR could be calculated.
the brain data on cerebellum are used, because it is known that in cerebellum there is no or there is very little deposition of amyloid beta.
d p
K b C
F
f 1 DVR
variables of the aggregation model
f = CF / (CF + CNS) - fraction of radioligand free from nonspecific binding bP – number of monomers per motif Kd - dissociation constant for radioligand and specific binding site
known from literature
Validation of the model against available PET data measured for both healthy subjects and AD patients Healthy subjects AD patients Then we simulate DVR and SUVR using model for healthy subjects and AD patients and compare it with experimental data.
Black: J Cereb Blood Flow Metab. 2005 Nov;25(11):1528-47 Blue: Arch Neurol. 2011 May;68(5):644-9 Red: J Cereb Blood Flow Metab. 2005 Nov;25(11):1528-47 Black: Neurobiol Aging. 2010 Aug;31(8):1275- 83 Blue: Neurobiol Aging. 2010 Aug;31(8):1275- 83 + J Cereb Blood Flow Metab. 2005 Nov;25(11):1528-47 J Cereb Blood Flow Metab. 2005 Nov;25(11):1528-47. Neurobiol Aging. 2010 Aug;31(8):1275-83 + J Cereb Blood Flow Metab. 2005 Nov;25(11):1528-47
It is not fitting It is not fitting It is not fitting It is not fitting
Drug treatment simulations (Ab42 insoluble)
We simulate treatment of AD patient by decreasing of synthesis or increasing of degradation. For this simulation we take AD patient model where we change degradation constants of soluble Ab after 40 years. Synthesis inhibition: Black – 25% Blue – 50% Red – 100% Synthesis inhibition: Black – 25% Blue – 50% Red – 100% Synthesis inhibition: Black – 25% Blue – 50% Red – 100% Degradation increase: Black – x2 Blue – x10 Degradation increase: Black – x2 Blue – x10 Degradation increase: Black – x2 Blue – x10
Start of treatment: 40 years Start of treatment: 50 years Start of treatment: 65 years It was shown that formation of plaques begin when Ab42 insoluble level is more than 300 nM [Am J Pathol. 1998 Jun;152(6):1633-40].
Drug treatment simulations (Ab42 soluble)
We simulate treatment of AD patient by decreasing of synthesis or increasing of degradation. For this simulation we take AD patient model where we change degradation constants of soluble Ab after 40 years. Synthesis inhibition: Black – 25% Blue – 50% Red – 100% Synthesis inhibition: Black – 25% Blue – 50% Red – 100% Synthesis inhibition: Black – 25% Blue – 50% Red – 100% Degradation increase: Black – x2 Blue – x10 Degradation increase: Black – x2 Blue – x10 Degradation increase: Black – x2 Blue – x10
Start of treatment: 40 years Start of treatment: 50 years Start of treatment: 65 years
and AD patients has been developed.
multiple years with 50% inhibition.
and current clinical trials due to insufficient duration of treatment and/or too little potency.
reasonable amount of time a compound will need to have the ability to reduce abeta production by >>50% (75-95% would be preferable).
Pfizer: ISBSPb:
DBSolve Optimum development team:
ISBSPb Gizzatkulov, Nail
DBSolve Optimum software package has been used to develop and analyze all the models presented.
Gizzatkulov N. et al; BMC Systems Biology, 2010, 4: 109 Timothy Nicholas, Hugh A. Barton, Yasong Lu, Sridhar Duvvuri Tatiana Karelina, Oleg Demin Jr , Sergey Belykh
http://www.insysbio.ru